A big data approach to cargo type prediction and its implications for oil trade estimation

To estimate global crude oil trade flows, current research either considers only crude oil tankers, or simply applies external information to distinguish between crude and refined product oil cargoes transported by coated product tankers; these limitations often reduce an estimation’s accuracy or compromise replicability. Our methodology directly addresses these issues by applying the random forest (RF) ensemble learning technique to Automatic Identification System (AIS) data in order to predict the cargo types of coated product tankers. By leveraging domain knowledge, we construct a set of unique input variables for the RF model, and use its predictions to quantify the global crude oil trade in a more accurate manner. Our estimation shows that coated product tankers were responsible for approximately 8% of global seaborne crude oil trade from 2017–2020. Further, unanticipated variations in the crude oil volume carried by these tankers are consistent with several major historical oil trade disruptions. Our study further extends current applications of AIS data in the domains of operations management and maritime transportation, and facilitates the exploration of the more minute characteristics of oil transportation. The resulting shipping dataset and associated decomposition strategy also enable relevant stakeholders to quickly identify emerging trade flow risks and adapt more effectively.